- I clarify about media. I agree about polls. I digress. On Tue, 05 Mar 2002 13:11:55 -0800, Jay Tanzman <[EMAIL PROTECTED]> wrote: > Rich Ulrich wrote: > > > Personally, I'm used to seeing 'margin of error' written and > > described very poorly or ambiguously, mainly in newspaper > > articles, where it is impossible to tell what it refers to. Unless > > the numbers are there so I can do the math. > > In reporting survey results to the public, if there is more than one result > reported, the single, so-called, margin of error, is usually the width of the > 95% confidence interval for the sample proportion 0.5. Since, for a given > sample size, any other sample proportion will have a narrower confidence > interval, the so-called margin of error is conservative. > > -Jay
All right. The Pittsburgh paper and the NY Times are both steady reporters of interesting results from Nature, Science, Lancet, NE J of M, and so on. For whatever reason, those clinical research reports are often unclear about whatever limits being reported. That's where I start looking for Ns, and making Poisson assumptions. By contrast, newspaper reports of media polls and surveys almost always meet live up to some standards that were approved by professional associations a few decades ago. I know there is a society of pollsters. I think that a journalism group also chastises offenders. How well do standards work on polls? - Polls are presented much better than they once were. They usually *do* tell the N, how it was selected, and what the exact questions were. - However, folks who follow media as a hobby are well-aware that Newt Gingrich and his pollster screwed up, on the "Contract with America" hype. Thus, you could say standards don't work, because offenses occurred, with relatively small attention. Or, the standards worked, because those folks won't repeat it, after their tons of negative publicity. (... breeding shame that led Newt to drop out. No doubt.) -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . ================================================================= . .
